CN102722759A

CN102722759A - Method for predicting power supply reliability of power grid based on BP neural network

Info

Publication number: CN102722759A
Application number: CN2012101560444A
Authority: CN
Inventors: 卫志农; 刘亚南; 孙国强; 孙永辉; 韦延方; 杨雄; 袁阳; 陆子刚; 王越; 陈婷; 杨友情; 江龙才; 吴常胜; 钱瑛; 周军; 李进
Original assignee: Hohai University HHU; Chizhou Power Supply Co of State Grid Anhui Electric Power Co Ltd
Current assignee: Hohai University HHU; Chizhou Power Supply Co of State Grid Anhui Electric Power Co Ltd
Priority date: 2012-05-17
Filing date: 2012-05-17
Publication date: 2012-10-10
Anticipated expiration: 2032-05-17
Also published as: CN102722759B

Abstract

The invention discloses a BP neural network-based power supply reliability prediction method. The present invention utilizes BP neural network to carry out modeling and forecasting, takes factors such as power grid supply capacity and natural disasters faced by the power grid as input variables of the model, and takes power grid failure and outage time that affects power supply reliability as the output of the model, trains and establishes a reliable power grid On this basis, principal component analysis is used to reduce the dimensionality of the input samples, which simplifies the input structure of the artificial neural network and improves the learning rate and accuracy of the network.

Description

Power Grid Power Supply Reliability Prediction Method Based on BP Neural Network

技术领域 technical field

本发明涉及一种电网供电可靠性预测方法，属于电力系统技术领域。The invention relates to a power grid power supply reliability prediction method, which belongs to the technical field of power systems.

背景技术 Background technique

供电可靠性是指供电系统对用户连续供电的能力。随着社会的发展和人民生活水平的提高，社会对供电可靠性的要求也越来越高，提高供电可靠性既是用户的需要，也是供电企业自身发展所追求的目标。近年来，国内外大电网发生的大面积长时间停电事故，不但造成巨大的经济损失，而且危及社会秩序。因此定量地评价和改善电力系统的供电可靠性，对可靠性的研究显得更加必要和迫切，可以说，电力系统可靠性的形成和发展是电力工业本身发展的客观规律所决定的。同时其它工业部门在研究和应用可靠性技术方面取得了积极成果，给电力系统供电可靠性的发展起到巨大的推动作用。Power supply reliability refers to the ability of the power supply system to continuously supply power to users. With the development of society and the improvement of people's living standards, the society has higher and higher requirements for the reliability of power supply. Improving the reliability of power supply is not only the need of users, but also the goal pursued by power supply enterprises. In recent years, large-scale and long-term power outages in large power grids at home and abroad have not only caused huge economic losses, but also endangered social order. Therefore, it is more necessary and urgent to quantitatively evaluate and improve the power supply reliability of the power system. It can be said that the formation and development of the reliability of the power system are determined by the objective law of the development of the power industry itself. At the same time, other industrial sectors have achieved positive results in the research and application of reliability technology, which has played a huge role in promoting the development of power supply reliability in power systems.

随着“十一五”规划接近尾声，电网公司希望通过科学建立可靠性预测模型来总结评估工作，同时更有效果的制定发展规划和目标。尤其通过加大电网投资，来达到提过供电可靠性的目的，否则投资变得没有意义。同时有必要分析各个因素对供电可靠性的影响，使得电网公司的投资可以根据不同的影响程度相应的增加或减少，对投资进行一定的规划，具有实际的意义。As the "Eleventh Five-Year Plan" draws to a close, power grid companies hope to summarize and evaluate work through the scientific establishment of reliability prediction models, and at the same time formulate development plans and goals more effectively. Especially by increasing the investment in the power grid to achieve the purpose of improving the reliability of power supply, otherwise the investment becomes meaningless. At the same time, it is necessary to analyze the impact of various factors on the reliability of power supply, so that the investment of the power grid company can be increased or decreased according to the degree of influence, and it is of practical significance to plan the investment.

传统的可靠性预测方法需要准确的电网结构和多年的元件可靠性指标历史数据，然而电网结构复杂，数据量很大，再加上目前无法确定目标年的具体网络结构，因此无法采用传统方法对整个电网的供电可靠性进行预测。Traditional reliability prediction methods require accurate power grid structure and historical data of component reliability indicators for many years. However, the power grid structure is complex and the amount of data is large. In addition, the specific network structure of the target year cannot be determined at present, so traditional methods cannot be used to The power supply reliability of the entire grid is predicted.

发明内容 Contents of the invention

本发明所要解决的技术问题在于克服现有技术不足，提供一种基于BP神经网络的电网供电可靠性预测方法，利用BP神经网络良好的非线性函数逼近能力，改善了预测模型的精度和泛化能力。The technical problem to be solved by the present invention is to overcome the deficiencies in the prior art and provide a method for predicting the reliability of power supply of power grid based on BP neural network, which improves the accuracy and generalization of the prediction model by utilizing the good nonlinear function approximation ability of BP neural network ability.

本发明具体采用以下技术方案解决上述技术问题。The present invention specifically adopts the following technical solutions to solve the above-mentioned technical problems.

基于BP神经网络的电网供电可靠性预测方法，包括以下步骤：A method for predicting reliability of power grid power supply based on BP neural network, including the following steps:

步骤A、分析并选取电网供电可靠性的影响因素；Step A, analyzing and selecting factors affecting the reliability of power grid power supply;

步骤B、根据所选取的影响因素的历史数据生成输入向量，以所对应的平均故障停电时间的历史数据作为输出，得到训练样本；Step B. Generate an input vector according to the historical data of the selected influencing factors, and use the corresponding historical data of the average fault outage time as output to obtain a training sample;

步骤C、利用所述训练样本对BP神经网络进行训练，得到训练后的BP神经网络；所述神经网络的输入层节点数为所述输入向量的特征维数，输出层节点数为1；Step C, using the training sample to train the BP neural network to obtain the trained BP neural network; the number of input layer nodes of the neural network is the characteristic dimension of the input vector, and the number of output layer nodes is 1;

步骤D、根据所选取的影响因素的待预测时刻的实际数据生成测试输入向量，并将测试输入向量输入训练后的BP神经网络，其输出即为待预测时刻的平均故障停电时间的预测值。Step D. Generate a test input vector according to the actual data of the selected influencing factors at the time to be predicted, and input the test input vector into the trained BP neural network, and its output is the predicted value of the average fault outage time at the time to be predicted.

上述技术方案中，所述输入向量可由各影响因素的数据直接生成，但当所选取的影响因素较多时，过多的变量会导致计算复杂度高，从而影响预测效率。为此，作为本发明的进一步改进方案，所述根据所选取的影响因素的历史数据生成输入向量，具体按照以下方法：首先，利用所选取的影响因素的历史数据构造特征向量；然后，对该特征向量进行降维处理，得到输入向量；所述根据所选取的影响因素的待预测时刻的实际数据生成测试输入向量，具体按照以下方法：首先，利用所选取的影响因素的待预测时刻的实际数据构造测试特征向量；然后，对该测试特征向量进行降维处理，得到测试输入向量。即通过对原始数据样本进行降维处理（特征提取）来降低预测的计算复杂度，提高效率。In the above technical solution, the input vector can be directly generated from the data of each influencing factor, but when there are many influencing factors selected, too many variables will lead to high computational complexity, thus affecting the prediction efficiency. For this reason, as a further improvement of the present invention, the generating of the input vector according to the historical data of the selected influencing factors specifically follows the following method: first, utilize the historical data of the selected influencing factors to construct a feature vector; then, for the The feature vector is subjected to dimensionality reduction processing to obtain the input vector; the test input vector is generated according to the actual data of the selected influencing factors at the time to be predicted, specifically according to the following method: first, using the actual data at the time to be predicted of the selected influencing factors The test feature vector is constructed from the data; then, the dimensionality reduction process is performed on the test feature vector to obtain the test input vector. That is to reduce the computational complexity of prediction and improve efficiency by performing dimensionality reduction processing (feature extraction) on the original data samples.

优选地，所述降维处理采用主成分分析（Principal ComponentAnalysis，PCA）方法。Preferably, the dimensionality reduction process adopts a principal component analysis (Principal Component Analysis, PCA) method.

优选地，所述影响因素包括：220kV架空线可用系数、110kV架空线可用系数、220kV变压器可用系数、110kV变压器可用系数、220kV断路器可用系数、110kV断路器可用系数、220kV容载比、110kV容载比、220kV供电半径、110kV供电半径、单位增供负荷220kV变电容量、单位增供负荷110kV变电容量、拉限电损失电量、农村供电可靠率、城市供电可靠率、雷击日数、暴雨日数、风灾日数、高温日数。Preferably, the influencing factors include: 220kV overhead line availability factor, 110kV overhead line availability factor, 220kV transformer availability factor, 110kV transformer availability factor, 220kV circuit breaker availability factor, 110kV circuit breaker availability factor, 220kV capacity-to-load ratio, 110kV capacity Load ratio, 220kV power supply radius, 110kV power supply radius, 220kV substation capacity per unit additional supply load, 110kV substation capacity per unit additional supply load, power loss due to power rationing, rural power supply reliability rate, urban power supply reliability rate, number of days of lightning strikes, and number of days of heavy rain , the number of wind disaster days, and the number of high temperature days.

本发明的电网供电可靠性预测方法利用BP神经网络进行建模预测，利用其良好的非线性函数逼近能力，改善预测模型的精度和泛化能力；并在此基础上使用主成分分析对输入样本进行降维处理，降低了算法复杂度，提高了预测效率。The power grid power supply reliability prediction method of the present invention uses BP neural network to carry out modeling prediction, utilizes its good nonlinear function approximation ability, improves the precision and the generalization ability of prediction model; And uses principal component analysis on this basis to input sample Dimension reduction processing reduces the complexity of the algorithm and improves the prediction efficiency.

附图说明 Description of drawings

图1为本发明预测模型的结构示意图；Fig. 1 is the structural representation of prediction model of the present invention;

图2为本发明方法中BP神经网络训练过程的流程图。Fig. 2 is the flowchart of BP neural network training process in the method of the present invention.

具体实施方式 Detailed ways

下面结合附图对本发明的技术方案进行详细说明：The technical scheme of the present invention is described in detail below in conjunction with accompanying drawing:

本发明的思路是利用人工智能神经网络构建一个电网可靠性预测模型，利用该预测模型对待预测时刻的电网可靠性进行预测。下面对该预测模型的建立过程进行详细说明。The idea of the present invention is to use artificial intelligence neural network to build a power grid reliability prediction model, and use the prediction model to predict the power grid reliability at the time to be predicted. The process of establishing the prediction model will be described in detail below.

输入变量的选择：Selection of input variables:

由于电网的供电可靠性主要与电网自身的供应能力和电网所处的自然环境密不可分，因此，本发明选取了体现电网自身供应能力的电网设备可用系数、供电可靠率、供电半径、容载比、拉限电损失电量、单位增供负荷新增变电容量，同时加入气象因素。具体的有19个因素：220kV和110kV架空线可用系数、220kV和110kV变压器可用系数、220kV和110kV断路器可用系数、220kV和110kV容载比、220kV和110kV供电半径、单位增供负荷220kV和110kV变电容量、农村和城市供电可靠率、雷击日数、暴雨日数、风灾日数和高温日数。Since the power supply reliability of the power grid is mainly inseparable from the power supply capacity of the power grid itself and the natural environment in which the power grid is located, the present invention selects the availability coefficient of power grid equipment, power supply reliability rate, power supply radius, and capacity-load ratio that reflect the power grid's own supply capacity. , power loss due to power rationing, new capacity for transformation per unit of additional supply load, and meteorological factors. Specifically, there are 19 factors: 220kV and 110kV overhead line availability factor, 220kV and 110kV transformer availability factor, 220kV and 110kV circuit breaker availability factor, 220kV and 110kV capacity-to-load ratio, 220kV and 110kV power supply radius, unit additional supply load 220kV and 110kV Substation capacity, rural and urban power supply reliability, lightning strike days, rainstorm days, wind disaster days and high temperature days.

输出变量的选择：Selection of output variables:

在日常生产中，影响供电可靠性的原因主要又两个方面：一是故障停电，二是预安排停电。电网系统各种因素对电网产生的最终直观表现为电网停电事故，因此本发明以年平均故障停电时间作为电网供电可靠性预测模型的输出量。In daily production, there are two main reasons affecting the reliability of power supply: one is a power outage due to a fault, and the other is a pre-arranged power outage. The final intuitive performance of various factors of the power grid system on the power grid is the power grid blackout accident, so the present invention uses the annual average fault power outage time as the output of the grid power supply reliability prediction model.

人工神经网络是人工智能技术的一种，具有大规模分布式并行处理、非线性、自组织、自学习、联想记忆等优良特性，因此可以作为一种预测手段。BP算法是训练人工神经网络的基本方法，其基本思想是最小二乘算法。它采用梯度搜索技术，以期使网络的实际输出值与期望输出值的误差均方值为最小。x₁,x₂,…,x_k为BP神经网络的输入变量，d₁,d₂,…,d_n为BP神经网络的预测值，w_ij为输入层与隐含层的权值，w_jk为隐含层与输出层的权值。输入节点为n，隐含层节点可设为l=2n，输出节点为m。其训练的具体步骤如下：Artificial neural network is a kind of artificial intelligence technology, which has excellent characteristics such as large-scale distributed parallel processing, nonlinearity, self-organization, self-learning, associative memory, etc., so it can be used as a means of prediction. BP algorithm is the basic method of training artificial neural network, and its basic idea is the least squares algorithm. It uses gradient search technology to minimize the mean square value of the error between the actual output value and the expected output value of the network. x ₁ , x ₂ ,…,x _k are the input variables of the BP neural network, d ₁ , d ₂ ,…,d _n are the predicted values of the BP neural network, w _ij are the weights of the input layer and the hidden layer, w _jk is the weight of the hidden layer and the output layer. The input node is n, the hidden layer node can be set to l=2n, and the output node is m. The specific steps of its training are as follows:

(1)BP神经网络初始化：根据模型确定输入层节点数、隐含层节点数和输出层节点数，初始化w_ij、w_jk、隐含层阀值a、输出层阀值b，给定学习速率和神经元激励函数；(1) BP neural network initialization: Determine the number of input layer nodes, hidden layer nodes and output layer nodes according to the model, initialize w _ij , w _jk , hidden layer threshold a, output layer threshold b, given learning rate and neuron activation functions;

（2）隐含层输出计算：根据输入向量，输入层和隐含层连接权值w_ij以及隐含层阀值a，计算隐含层输出H：(2) Hidden layer output calculation: According to the input vector, input layer and hidden layer connection weight w _ij and hidden layer threshold a, calculate the hidden layer output H:

$H h = = f f (({Σ Σ}_{i i = = 11}^{k k} {w w}_{ij ij} {x x}_{i i} - - {a a}_{j j})) j j = = 1,2 1,2,, \cdot &Center Dot; \cdot &Center Dot; \cdot &Center Dot;,, l l$

式中：l为隐含层节点数；f为隐含层激励函数，本发明选函数为：In the formula: l is the number of hidden layer nodes; f is the hidden layer activation function, and the selection function of the present invention is:

$f f ((x x)) = = \frac{11}{11 + + {e e}^{- - x x}}$

（3）输出层输出计算：根据隐含层输出H，连接权值w_jk和阀值b，计算BP神经网络预测输出O：(3) Output layer output calculation: According to the hidden layer output H, connect the weight w _jk and the threshold b, calculate the BP neural network prediction output O:

${O o}_{k k} = = {Σ Σ}_{j j = = 11}^{m m} {H h}_{j j} {w w}_{jk jk} - - {b b}_{k k} k k = = 1,2 1,2,, \cdot &Center Dot; \cdot &Center Dot; \cdot &Center Dot;,, m m$

（4）误差计算：根据网络预测输出O和期望输出d，计算网络预测误差e：(4) Error calculation: Calculate the network prediction error e according to the network prediction output O and the expected output d:

e_k=d_k-O_k k=1,2,…,me _k =d _k -O _k k=1,2,...,m

（5）权值更新：根据网络预测误差e更新网络连接权值w_ij和w_jk：(5) Weight update: update the network connection weights w _ij and w _jk according to the network prediction error e:

${w w}_{ij ij} = = {w w}_{ij ij} + + η η {H h}_{j j} ((11 - - {H h}_{j j})) x x ((i i)) {Σ Σ}_{k k = = 11}^{m m} {w w}_{jk jk} {e e}_{k k} i i = = 1,2 1,2,, \cdot \cdot \cdot \cdot \cdot &Center Dot;,, n no;; j j = = 1,2 1,2,, \cdot &Center Dot; \cdot \cdot \cdot \cdot,, l l$

w_jk=w_jk+ηH_je_k j=1,2,…,l;k=1,2,…,mw _jk =w _jk +ηH _j e _k j=1,2,…,l;k=1,2,…,m

式中：η为学习速率，0≤η≤1；In the formula: η is the learning rate, 0≤η≤1;

（6）阀值更新：根据网络预测误差e更新网络节点阀值a，b：(6) Threshold update: update the network node threshold a, b according to the network prediction error e:

${a a}_{j j} = = {a a}_{j j} + + η η {H h}_{j j} ((11 - - {H h}_{j j})) {Σ Σ}_{k k = = 11}^{m m} {w w}_{jk jk} {e e}_{k k} j j = = 1,2 1,2,, \cdot \cdot \cdot \cdot \cdot \cdot,, l l$

b_k=b_k+e_k k=1,2,…,mb _k =b _k +e _k k=1,2,...,m

（7）判断算法迭代是否结束，若没有介绍，返回步骤（2）。(7) Determine whether the algorithm iteration is over, if there is no introduction, return to step (2).

由于本发明选择了19个影响因素作为输入变量，因此可将BP神经网络的输入层节点设为19，输出层节点数设为1。但采用该方案时，较多变量会导致计算复杂度高，影响预测效率。因此，本发明对输入变量先进行降维处理。可采用的降维方法较多，例如可以采用现有的主成分分析、非负矩阵分解、粗糙集约简和灰色关联度等方法。本发明优选采用主成分分析方法。其具体过程如下：Since the present invention selects 19 influencing factors as input variables, the input layer nodes of the BP neural network can be set to 19, and the number of output layer nodes can be set to 1. However, when using this scheme, more variables will lead to high computational complexity and affect the prediction efficiency. Therefore, the present invention performs dimensionality reduction processing on input variables first. There are many dimension reduction methods that can be used, such as the existing principal component analysis, non-negative matrix decomposition, rough set reduction and gray relational degree and other methods. The present invention preferably adopts the principal component analysis method. The specific process is as follows:

（1）原始数据标准化：本发明的输入变量由电网设备可用系数、供电可靠率、供电半径、容载比、拉限电损失电量、单位增供负荷新增变电容量、高温日数、雷击日数、风灾日数和暴雨日数组成，输出变量是年平均故障停电时间。原始数据组成的矩阵记为(1) Raw data standardization: The input variables of the present invention are composed of grid equipment availability factor, power supply reliability rate, power supply radius, capacity-load ratio, power loss due to power cuts, new capacity of transformers added per supply increase load, high temperature days, and lightning strike days , the number of wind disaster days and the number of rainstorm days, and the output variable is the annual average fault power outage time. The matrix composed of the original data is denoted as

$X x = = [\begin{matrix} {x x}_{1111} & {x x}_{1212} & \cdot \cdot \cdot \cdot \cdot \cdot & {x x}_{11 p p} \\ {x x}_{21 twenty one} & {x x}_{22 twenty two} & \cdot \cdot \cdot &Center Dot; \cdot &Center Dot; & {x x}_{22 p p} \\ \cdot \cdot & \cdot \cdot & \cdot \cdot & \cdot \cdot \\ \cdot \cdot & \cdot \cdot & \cdot \cdot & \cdot \cdot \\ \cdot \cdot & \cdot \cdot & \cdot \cdot & \cdot \cdot \\ {x x}_{n no 11} & {x x}_{n no 22} & \cdot \cdot \cdot \cdot \cdot \cdot & {x x}_{np np} \end{matrix}]$

式中：n是样本个数，p是每个样本的维数，x_ij为第i个样本的第j维取值。用x₁,x₂,…,x_p分别表示矩阵X的各列矢量，有In the formula: n is the number of samples, p is the dimension of each sample, and x _ij is the value of the jth dimension of the i-th sample. Use x ₁ , x ₂ ,…, x _p to denote the column vectors of matrix X respectively, we have

${x x}_{j j}^{* *} = = \frac{{x x}_{j j} - - E E. (({x x}_{j j}))}{\sqrt{Var Var (({x x}_{j j}))}} ((j j = = 1,2 1,2,, \cdot &Center Dot; \cdot &Center Dot; \cdot &Center Dot;,, p p))$

式中：E(x_j)和Var(x_j)分别表示x_j的均值和方差。In the formula: E(x _j ) and Var(x _j ) represent the mean and variance of x _j respectively.

（2）计算样本各维间相关系数矩阵R=(r_ij)_p×p：(2) Calculate the correlation coefficient matrix R=(r _ij ) _p×p between the dimensions of the sample:

${r r}_{ij ij} = = \frac{Cov Cov (({x x}_{i i},, {x x}_{j j}))}{\sqrt{Var Var (({x x}_{i i})) Var Var (({x x}_{j j}))}} = = \frac{11}{n no} {Σ Σ}_{k k = = 11}^{n no} {x x}_{ki the ki}^{* *} {x x}_{kj kj}^{* *} ((i i,, j j = = 1,2 1,2,, \cdot \cdot \cdot &Center Dot; \cdot &Center Dot;,, p p))$

式中：Cov(x_i,x_j)表示原始数据矩阵中第i列与第j列之间的协方差。Where: Cov( _xi , x _j ) represents the covariance between column i and column j in the original data matrix.

上式等价于The above formula is equivalent to

$R R = = \frac{11}{N N - - 11} {(({X x}^{* *}))}^{T T} (({X x}^{* *}))$

式中：X^*是将X标准化后的数据矩阵，可以看出，R为半正定对称矩阵。In the formula: X ^* is the data matrix after X is standardized. It can be seen that R is a positive semi-definite symmetric matrix.

（3）求R的特征值为λ₁≥λ₂≥…≥λ_p和其对应的单位化特征向量μ₁,μ₂,…,μ_p。(3) Find the eigenvalues of R λ ₁ ≥λ ₂ ≥…≥λ _p and their corresponding unitized eigenvectors μ ₁ , μ ₂ ,…,μ _p .

（4）确定主成分个数。分别计算方差贡献率η_i和累计方差贡献率β_i (4) Determine the number of principal components. Calculate variance contribution rate η _i and cumulative variance contribution rate β _i respectively

${η η}_{i i} = = {λ λ}_{i i} / / {Σ Σ}_{i i = = 11}^{p p} {λ λ}_{i i} \times \times 100100 % %$

${β β}_{i i} = = {Σ Σ}_{k k = = 11}^{i i} {λ λ}_{k k} / / {Σ Σ}_{k k = = 11}^{p p} {λ λ}_{k k} \times \times 100100 % %$

选取主成分的个数取决于累计方差贡献率，通常累计方差贡献率大于85%-90%，对应的前k个主成分便包含p个原始变量所能提供的绝大部分信息，则主成分个数就是k个，从而实现了R^p →R^k的线性变换，达到特征提取和降维的目的。The number of selected principal components depends on the cumulative variance contribution rate, usually the cumulative variance contribution rate is greater than 85%-90%, and the corresponding first k principal components contain most of the information that p original variables can provide, then the principal components The number is k, which realizes the linear transformation of R ^p → R ^k , and achieves the purpose of feature extraction and dimensionality reduction.

本发明完整的预测模型结构如图1所示。图2显示了BP神经网络的训练过程。The complete prediction model structure of the present invention is shown in FIG. 1 . Figure 2 shows the training process of the BP neural network.

本发明在人工神经网络的基础上引入主成分分析法对样本进行特征提取，消除变量间的相关性再进行建模，这样既结合了主成分分析（PCA）的特征提取能力，又利用了人工神经网络良好的非线性函数逼近能力，从而改善了预测模型的精度和泛化能力。The present invention introduces the principal component analysis method on the basis of the artificial neural network to extract the features of the sample, eliminates the correlation between variables and then performs modeling, which not only combines the feature extraction ability of the principal component analysis (PCA), but also utilizes artificial The good nonlinear function approximation ability of the neural network improves the accuracy and generalization ability of the prediction model.

为了验证本发明方法的有效性，进行以下实验：以华东某市2003年-2008年19个影响因素和年平均故障停电时间作为模型的训练样本，以2009年的实际数据作为测试样本，在Matlab环境下训练并建立该地区电网供电可靠性预测模型。首先利用主成分分析方法对2003年-2008年19个影响因素的历史数据进行降维处理。包括以下步骤：In order to verify the effectiveness of the method of the present invention, carry out following experiment: take 2003-2008 year 19 influence factors and the annual mean power outage time of a certain city in East China as the training sample of model, take the actual data in 2009 as test sample, in Matlab Under the environment, train and establish the reliability prediction model of power grid in this area. Firstly, the historical data of 19 influencing factors from 2003 to 2008 are processed by principal component analysis for dimensionality reduction. Include the following steps:

步骤1：将输入变量标准化处理，计算相关系数矩阵。Step 1: Standardize the input variables and calculate the correlation coefficient matrix.

步骤2：由相关系数矩阵计算特征值、各个主成分的贡献率以及累计贡献率，具体见表1（这里只列举7个主成分）。Step 2: Calculate the eigenvalues, the contribution rate of each principal component and the cumulative contribution rate from the correlation coefficient matrix, see Table 1 for details (only 7 principal components are listed here).

表1特征值及主成分贡献率Table 1 Eigenvalues and principal component contribution rates

由表1可知，前5个主成分的累计贡献率已高达96.99%（大于95%），说明前5个主成分提供了原始数据比较充足的信息，因此提取5个主成分进行预测，对于5个特征值分别求出其特征向量，再计算各对于5个特征值分别求出其特征向量，再计算各变量在主成分上的载荷，具体的计算结果见表2。It can be seen from Table 1 that the cumulative contribution rate of the first 5 principal components has reached 96.99% (greater than 95%), indicating that the first 5 principal components provide sufficient information for the original data, so 5 principal components are extracted for prediction. Calculate the eigenvectors of the five eigenvalues respectively, and then calculate the eigenvectors for each of the five eigenvalues, and then calculate the loads of each variable on the principal components. The specific calculation results are shown in Table 2.

表2主成分负载Table 2 Principal component loadings

将前5个主成分构成的新样本空间作为BP神经网络的输入量进行训练和预测，BP神经网络初始学习率设为0.1，输入层节点数为5，隐含层节点数为11，输出层节点数为1。预测结果见表3，其中“BP”是指不采用降维处理，直接以传统BP神经网络进行预测；“PCA_BP”是指本具体实施方式中所述方法。The new sample space composed of the first five principal components is used as the input of the BP neural network for training and prediction. The initial learning rate of the BP neural network is set to 0.1, the number of nodes in the input layer is 5, the number of nodes in the hidden layer is 11, and the number of nodes in the output layer is The number of nodes is 1. The prediction results are shown in Table 3, where "BP" refers to the traditional BP neural network for prediction without dimension reduction processing; "PCA_BP" refers to the method described in this specific embodiment.

表3预测结果分析Table 3 Analysis of prediction results

从表3可以看出，采用PCA方法对神经网络的输入进行降维，不但减少了输入变量维数，降低了计算复杂度，而且提高了预测精度。It can be seen from Table 3 that using the PCA method to reduce the dimensionality of the input of the neural network not only reduces the dimension of the input variables, reduces the computational complexity, but also improves the prediction accuracy.

Claims

1. based on the mains supply reliability prediction method of BP neural network, it is characterized in that, may further comprise the steps:

Steps A, analyze and choose the influence factor of mains supply reliability;

Step B, generate input vector, as output, obtain training sample with the historical data of pairing mean failure rate power off time according to the historical data of selected influence factor;

Step C, utilize said training sample that the BP neural network is trained, the BP neural network after obtaining training; The input layer number of said neural network is the intrinsic dimensionality of said input vector, and output layer node number is 1;

Step D, generate the test input vector according to the real data in moment to be predicted of selected influence factor, and will test the BP neural network after the input vector input is trained, its output is the predicted value of the mean failure rate power off time in the moment to be predicted.

2. according to claim 1 based on the mains supply reliability prediction method of BP neural network, it is characterized in that,

Said historical data according to selected influence factor generates input vector, specifically according to following method: at first, utilize the historical data structural attitude vector of selected influence factor; Then, this proper vector is carried out dimension-reduction treatment, obtain input vector;

The real data in the said moment to be predicted according to selected influence factor generates the test input vector, specifically according to following method: at first, utilize the real data structure testing feature vector in the moment to be predicted of selected influence factor; Then, this testing feature vector is carried out dimension-reduction treatment, obtain testing input vector.

3. like the said mains supply reliability prediction method of claim 2, it is characterized in that principal component analytical method is adopted in said dimension-reduction treatment based on the BP neural network.

4. like the said mains supply reliability prediction method of claim 3 based on the BP neural network; It is characterized in that; When adopting principal component analytical method to carry out dimensionality reduction, choose the major component of accumulative total variance contribution ratio greater than a predetermined threshold value, the span of said threshold value is 85%-90%.

5. like each said mains supply reliability prediction method of claim 1-4 based on the BP neural network; It is characterized in that said influence factor comprises: 220kV pole line availability coefficient, 110kV pole line availability coefficient, 220kV transformer availability coefficient, 110kV transformer availability coefficient, 220kV isolating switch availability coefficient, 110kV isolating switch availability coefficient, 220kV hold carry than, 110kV hold carry ratio, 220kV radius of electricity supply, 110kV radius of electricity supply, unit increase supply load 220kV power transformation capacity, unit to increase to supply load 110kV power transformation capacity, draw the loss electric weight of rationing the power supply, rural power service reliability, urban electricity supply reliability, number of days, heavy rain number of days, disaster caused by a windstorm number of days, high temperature number of days are struck by lightning.